Speech-to-Text Built for African Languages

Turn recorded speech into transcripts and SRT subtitles through the dashboard or API.

Competitive speech to text accuracy across African and Global context

<30s

Latency on a 30-minute audio file

Supported African and global languages

Hear the input. Inspect the result.

Perigee-1-transcribe

+ 7 more languages supported

Morning news bulletin

0:000:11

To sai a gyara zama, don sauraron cikakkun labarai. Rundunar sojin Najeriya ta kara kaimi wajen aikin tabbatar da tsaro cikin gida a sassan jihohin Kwara da Naija da sauran fadin kasar

Built for recordings that need to be text.

Turn broadcast media, customer calls, and field interviews into structured transcripts and subtitles.

Media & broadcasting

Generate accurate transcripts and timed SRT subtitles for broadcasts.

Hausa broadcast recording

Rundunar sojin Najeriya ta kara kaimi wajen aikin tabbatar da tsaro cikin gida a sassan jihohin Kwara da Naija.

Generated transcript & SRT subtitles

00:00:04,500 --> 00:00:10,000 Rundunar sojin Najeriya ta kara kaimi wajen aikin tabbatar da tsaro cikin gida.

See how Perigee-1-transcribe performs

Objective Word Error Rate (WER) evaluations on verified speech corpora. Lower WER indicates higher transcription accuracy.

Dataset:NaijaVoices•8 models evaluated

Lower WER = higher accuracyPerigee-1-transcribeComparison Models

OrbitalsAI (Perigee-1-transcribe)

12.60%

Sahara V2

16.73%

Gemini 3.0 Flash

29.13%

MMS-1B

39.37%

Gemini 2.0 Flash

52.16%

Whisper Large-v3

96.99%

GPT-4o audio-preview

119.74%

Whisper Medium

> 120%

One request. A result when it is ready.

File transcription is asynchronous. Submit audio once, then check the job until processing is complete.

Preview status flowTask 4821

Upload receivedAudio stored and validated
Job acceptedTask ID returned to your application
ProcessingPoll the status endpoint
Result readyRead the transcript or SRT

HTTP/1.1 202 Accepted
Content-Type: application/json

{
  "message": "Audio file uploaded successfully",
  "task_id": 4821,
  "status": "processing",
  "srt_requested": true
}

The transcription output, without guesswork.

Example completed resultinterview-hausa.mp3

Completed

Language: Hausa
Model: Perigee-1-transcribe
Duration: 18:42
Output: Text + SRT

The completed response includes transcript text and can include SRT content when requested.

Long recordingsProcessed in manageable audio segments.

Selected languageChoose the source language when submitting.

Job statusRetrieve processing state and completed output.

API authenticationUse an API key or authenticated dashboard session.

Speak in ten languages. Accurately transcribed.

Pre-trained on native speech distributions, tonal contours, and conversational code-switching across Africa.

Spoken languages

10 languages

Available results

2 results

Start with one recording.

Use the dashboard or connect Perigee-1-transcribe to your application.

Product updates

Receive release notes, product news, and practical guides.

By providing your email address, you consent to receiving our monthly newsletter. For more details, please review our privacy policy. You may withdraw your consent at any time.